Where Hybrid Workers Live Changes How Often They Stay Home

An open-access study published in the August 2026 issue of Cities compares regular hybrid workers in Paris, New York and London using harmonised 2023 survey data and an interpretable XGBoost model. It classifies working from home three to five days per week as high frequency and finds that commuting, housing, employment and demographic associations differ substantially among the metropolitan areas.

The clearest contrast concerns centrality: frequent home working was concentrated near New York’s central business district but became more likely with greater residential distance from central London. Paris had lower overall frequency and a narrower high-frequency profile. These are observational associations, not evidence that moving closer to or farther from a city centre changes an employee’s schedule.

Why one distance rule fails across the cities

A home’s distance from downtown and the length of a worker’s commute are related measures, but they are not interchangeable. Someone may live near the metropolitan core and work in an outer district, while another person living farther out may commute to a workplace that is also outside the centre. The model considered home and workplace geography alongside personal, housing and employment characteristics.

The analysis used XGBoost to capture curved relationships and interactions that can be missed by a single average effect. SHAP values were then used to make the model’s predictions interpretable, showing which variables contributed to higher or lower predicted frequency for observations in each metropolitan sample.

This approach can reveal that a factor matters differently at different points in its range. It does not turn a predictive pattern into a causal mechanism: employer rules, occupation, earnings, housing needs and residential choice may influence both where workers live and how many days they stay home.

London’s frequent home workers were more common farther out

London showed the strongest outward pattern. Within the population studied, living farther from the central business district was associated with a greater probability of belonging to the high-frequency home-working category.

A long journey offers one plausible interpretation: workers facing more demanding travel may value schedules that eliminate additional commutes. The reverse sequence is also plausible, however. Employees who already have substantial flexibility may be more willing to choose housing farther from central employment, and the survey cannot determine which decision came first.

The result therefore applies to regular hybrid workers with workplaces inside the study’s London metropolitan boundary. It should not be extended automatically to all British employees, fully remote workers or people whose workplaces fall outside that geography.

New York reverses London’s centrality pattern

New York did not reproduce London’s outward gradient. High-frequency home working was concentrated among people who both lived and worked near the central business district, according to the authors’ earlier SSRN version of the comparison.

This matters because a central residence is often treated as a proxy for easy office access and therefore more frequent attendance. The New York profile shows why that shortcut can fail: proximity to dense employment did not prevent regular hybrid workers in the sample from spending most of the working week at home.

The finding also separates downtown distance from the actual home-to-work journey. Two residents at similar distances from the centre can have different workplace locations, transport options and housing conditions. Those configurations may distinguish their schedules more effectively than a single measure of residential centrality.

Paris has a narrower high-frequency profile

Paris did not fit either the London or New York pattern. Home-working frequency was lower overall, the group working from home most of the week had a narrower profile, and spatial differentiation was weaker than in the other metropolitan samples.

That does not mean geography was irrelevant in Paris. It means distance and centrality provided less of the clear separation seen in the two other cities when combined with the available employment, housing and demographic variables. Institutional arrangements and the composition of eligible work may also differ, but the model does not isolate any single explanation as the cause.

The comparison is consequently asymmetric: London presents a relatively clear outward residential association, New York a central concentration involving both residence and workplace, and Paris a less spatially distinct high-frequency group. Paris should not be placed mechanically at an intermediate point on a single suburban-to-central scale.

Housing and demographics depend on metropolitan context

Income, gender and housing conditions were associated with home-working frequency, but their contribution varied with spatial and employment configurations. The study therefore does not support a universal demographic portrait of the frequent hybrid worker across all three cities.

Housing can affect whether sustained home working is practical, while income and gender can be connected to occupation, household responsibilities and access to flexible employment. Yet these variables interact with local labour markets and residential systems. An association that helps distinguish schedules in one city may contribute little to prediction in another.

For employers comparing attendance across offices, the immediate implication is methodological. Home location, workplace location, commute distance and metropolitan area need to remain separate variables; combining them into a company-wide “distance from downtown” measure could conceal opposite local patterns.

The study describes patterns rather than policy effects

The publication status is corroborated by the author’s HAL research record for the Cities article. The underlying respondent-level data are not publicly shareable, which limits independent reanalysis beyond the information and results released with the paper.

The findings do not estimate what would happen after an attendance mandate, an office relocation or a transport subsidy. Nor do they show that changing a worker’s address would alter home-working frequency. Answering those questions would require longitudinal observations or a design that evaluates a defined policy change.

As of publication, the supported conclusion is narrower but useful: metropolitan geography is associated with frequent home working in different, non-linear ways across Paris, New York and London. Repeated harmonised surveys would be needed to establish whether these city profiles persist as employer policies, housing choices and transport conditions evolve.

Also read:

• LLMs Explained: How Large Language Models Work (and Where They Commonly Fail)
• AI Job Disruption: Workers Most Exposed to AI Are Often Best Prepared to Adapt, Study Finds
• Health diagnostics for remote workers: how IT professionals and freelancers can detect body imbalances early